Lactancia materna

Table 2.3 reports for each treatment the mean and the standard deviation of the second mover’s e¤ort across individuals and all rounds.

Table 2.3: Mean and standard deviation of e2

Treatment Mean Standard Number of deviation observations T-NO 50.29 13.23 126 T-NO* 49.07 14.21 252 T-IMP 53.49 23.98 238 T-P 54.25 23.72 252 T-PPlus 55.26 18.34 210

*: e1 and e2 are considered.

In T-NO the second mover’s average e¤ort is around 50 (with or without the de- cisions of the …rst movers)30 which is close to our prediction of about 52.74. If, in contrast, an informative signal is available for the second mover, the second mover works more on average. This may be caused by the signal’s realized value the second

30_{As in T-NO the second mover does not receive a signal prior to his decision, the …rst mover’s}

decision problem is theoretically identical to the second mover’s decision problem. Therefore, both agents’decisions of T-NO may be considered in the analysis of the second mover’s behavior.

Figure 2.2: The second mover’s average reaction in T-P and T-PPlus 0 10 20 30 40 50 60 70 80

predicted for T-P and T- PPlus

observed in T-P observed in T-PPlus

A v e ra ge e ff ort of a ge nt 2

when agent 1's contribution is high when agent 1's contribution is low

mover observes in these treatments. In the following we, therefore, illustrate how the second mover reacts to the signals. The second mover’s average reaction to the …rst mover’s contribution is considerably di¤erent than predicted: In T-P the second mover works more when the …rst mover’s contribution is high rather than low, and in T- PPlus the second mover’s average e¤ort does not vary in the …rst mover’s contribution. Figure 2.2 reports the second mover’s average e¤ort conditional on the …rst mover’s contribution in T-P and T-PPlus.

If we consider the second mover’s reaction to the …rst mover’s e¤ort in T-IMP and T-PPlus, we …nd that the second mover’s e¤ort is positively correlated with the …rst mover’s e¤ort: In T-IMP and in T-PPlus the correlation coe¢ cient is equal to 0.24 and 0.12, respectively. The self-interest model, in contrast, predicts a negative correlation in T-IMP and no correlation in T-PPlus.

In order to analyze the second mover’s behavior more closely and to test our hy- potheses regarding the second mover, we regress the second mover’s e¤ort31 _{on treat-} ment dummies, the interactions of treatments dummies with available signals, and control variables such as sex and risk aversion32_{. In the …rst two columns of Table}

31_{In our regressions we also consider the …rst mover’s decisions in T-NO. Our results do not change}

considerably when we only consider the second mover’s decisions.

32_{We create the variable for risk aversion from our measure of risk attitude that we elicited in the}

2.4 we report the results of two random e¤ects panel regressions where one (Tobit) captures that our dependent variable is (weakly) between 0 and 80.

Table 2.4: Regression results on e2 and e1

Dependent variable: e2 e2 e1 e1

Panel (re) Panel (re) Panel (re) Panel (re) (Tobit) (Tobit) Intercept +47.30 +46.73 +45.58 +44.65 (0.000) (0.000) (0.000) (0.000) T-IMP - 09.72 - 13.17 +07.49 +07.72) (0.066) (0.045) (0.092) (0.124) T-P - 01.44 - 01.73 +02.54 +01.88 (0.737) (0.752) (0.568) (0.707) T-PPlus - 05.46 - 05.65 +10.66 +12.17 (0.432) (0.496) (0.020) (0.019) T-IMP * e1 +00.24 +00.34 (0.000) (0.000) T-PPlus * e1 +00.25 +00.27 (0.008) (0.014) T-P * b1 +12.56 +16.01 (0.000) (0.000) T-PPlus * b1 - 02.88 - 03.03 (0.275) (0.313) Sex +05.65 +08.05 +06.85 +08.54 (1 if male, 0 else) (0.059) (0.035) (0.040) (0.023) Risk aversion - 01.65 - 02.88 - 02.53 - 01.92 (1 if risk averse, 0 else) (0.582) (0.453) (0.454) (0.617)

Number of observations 952 952 952 952 (Pseudo) R-squared 0.069 0.089

(re): random e¤ects

In all four regressions e1 and e2 in T-NO are considered.

Numbers in brackets represent the p-values of the coe¢ cients.

When only the …rst mover’s e¤ort is observable (T-IMP), the second mover’s e¤ort signi…cantly increases in the …rst mover’s e¤ort. When only the …rst mover’s contri- bution is observable (T-P), the second mover’s e¤ort signi…cantly increases in the …rst mover’s contribution. When both the …rst mover’s e¤ort and contribution are observ- able (T-PPlus), the second mover’s e¤ort signi…cantly increases in the …rst mover’s e¤ort. We do, however, not …nd a signi…cant e¤ect of the …rst mover’s contribution on the second mover’s e¤ort in T-PPlus. These results con…rm our previous observa- tions from Figure 2.2 and from the correlation coe¢ cients in T-IMP and T-PPlus: In contrast to our predictions, the second mover positively reacts to informative signals

– in T-PPlus at least to the …rst mover’s e¤ort. The …rst mover’s e¤ort seems to be an especially stimulating signal as it even a¤ects the second mover’s e¤ort in T-PPlus. Consequently, we reject Hypotheses 2.1(i) and 2.1(ii). We also reject Hypothesis 2.3 since the second mover’s behavior in T-PPlus varies with the …rst mover’s e¤ort and, therefore, does not equal the second mover’s behavior in T-P.

We summarize our …ndings on the second mover’s behavior in the following three results:

Result 2.1: Higher average e¤ort of the second mover in T-IMP, T-P, and T- PPlus than in T-NO: On average the second mover exerts more e¤ort when informative signals are available for the second mover.

Result 2.2: Increasing reaction of the second mover in T-IMP, T-P, and T-PPlus: In contrast to Hypotheses 2.1(i) and 2.1(ii), the second mover exerts more e¤ort, the higher the …rst mover’s e¤ort in T-IMP and T-PPlus, or the higher the …rst mover’s contribution in T-P.

Result 2.3: E¤ect of the …rst mover’s e¤ort on the second mover’s behavior when the …rst mover’s contribution is observable: In contrast to Hypothesis 2.3, the second mover’s e¤ort increases in the …rst mover’s e¤ort in T-PPlus.